S5.4b Genomic variation structure in clinical isolate collections alters benefits of association screening: a study of Cryptococcus neoformans var. grubii in Colombia

S5.4 Free oral paper session, September 22, 2022, 3:00 PM - 4:30 PM Objective Recent genome-wide genotype-phenotype association studies (GWAS) of sets of clinical fungal isolates have reported problems resulting from strong phylogenetic clustering (apparent quasi-clonal population structure, difficulties in localizing association signals on chromosomes). We wished to directly investigate such difficulties, for whole-genome sequence data, and corresponding phenotypic data we had generated. Methods Our recently released Illumina read sequences (Bioproject PRJNA669191) for 29 clinical Colombian isolate genomes of Cryptococcus neoformans var. grubii (28 VNI and 1 VNII as outgroup)12 were assembled and aligned to the annotated H99 VNI reference genome. Single nucleotide polymorphisms (SNPs) were called for the VNI alignment. MLST types, phenotypes, and source metadata of the same isolates12 were integrated with the genomic data. Analyses were interpreted in the light of published literature. Results The 28 VNI isolates were assigned an order considering MLST classification and whole-genome phylogenetic distances. Most SNPs were biallelic, allowing straightforward barcoding: each of the 28 isolates was given a bar if it had the minor allele. Barcode profiles (SNP state vectors) were found repeated numerous times within genes and across entire chromosomes, indicating that neither fine nor ‘rough’ mapping of associations would succeed, except where flagged genes had already been characterized by relevant molecular biology experiments. The same repetitiveness can allow amplicons of individual MLST loci to tag and inform on variation present in other loci in the genome. The gene for the capsule-associated protein CAP59 and its promoter region illustrate such implications at a small (genic) scale. Conclusions Most associations of phenotypes with fungal genomes from modestly sampled clinical isolate populations cannot be localized to one or a few regions in a genome. This problem is less restrictive in human GWAS studies or (possibly) in studies of recombining fungi sampled from the environment3. It limits the benefits for molecular etiologic understanding that one might expect from GWAS of isolates obtained in clinical contexts, where key fungal phenotypes can be under stronger selection pressure. Explanatory power may be low if one compares just MLST types or phylogenetic clusters4 or examines only variation within individual MLST types. More informative results may be gained by restricting the genomic search space a priori to a subset of genes that are likely to be related to a phenotype, and then zooming in to individual SNPs that are known/suspected to affect protein structure/function1 or influence transcription/translation of the gene. Where the modes of regulation of a key gene are still not completely characterized despite decades of research, such as in CAP59, the presence of a phenotypically monitored, lone SNP in a strategic position within an almost SNP-less region can give potentially precious hints where otherwise none exist; we illustrate this strategy of SNP-focused local exploration for the promoter/enhancer and currently used MLST amplicon of CAP59. Sources